Oak wilt fungus infected over 266,000 oak trees between 2007 and 2016 in Minnesota, making it the second greatest invasive pathogen threat to the state behind Dutch elm disease. The fungus is found in more than a quarter of all Minnesota counties. Because the fungus spreads primarily by the tree’s roots, oak wilt can be difficult to detect. Monitoring forest health by foot has been expensive and time-consuming. But new technology like satellite imaging and spectral technology can analyze a tree’s vitals even before it shows physical signs of sickness.
One of the key challenges of detecting oak wilt is how similar the symptomatic bronze or brown wilting leaves can look to signs of drought stress and other diseases. Finding infected trees in a large plot of forest land can be like searching for a needle in a haystack, especially when wilting is in its more subtle early stages. Using handheld devices, drones and airborne hyperspectral sensors, UMN researchers in collaboration with the University of Wisconsin and University of Nebraska and with the help of NASA pilots have been capturing the photons — packets of energy — reflecting off forest canopies. This light contains information about the species of tree, its physiological condition and health, and stress it is under.
Using hyperspectral image data, the team can now identify oaks from other tree species with 95 percent accuracy and differentiate between healthy and trees infected with oak wilt with 84 percent accuracy.
The team also determined some best practices for using hyperspectral technology in this way, including identifying: the most important wavelengths to identify oak species, red oaks, and diseased red oaks; several multispectral indices associated with physiological decline that can detect differences between healthy and diseased trees; and the best time of year to conduct similar research (August). The wavelengths identified are also among the most important wavelengths for disease detection within PLS-DA models, showing a convergence of the methods.
This study highlights how hyperspectral models can differentiate oak wilt from other causes of tree decline, and that detection is correlated with biological mechanisms of oak wilt infection and disease progression. It has also shown how, within a canopy, symptom heterogeneity can reduce detection, but that symptomatic leaves and tree canopies are suitable for highly accurate diagnosis. The team’s next step, aimed at lowering cost, will be to test protocols that can use satellites or drones instead of manned flights for canopy detection.
Oaks make up nearly 30 percent of our North American forests and are one of the most important tree lineages in North America and the Northern Hemisphere. Not only do oaks store and filter carbon out of the atmosphere, oak trees are essential in preventing run-off, preserving habitat, and even fostering the economy through the export of lumber. Homeowners and municipalities are spending millions of dollars annually to treat, remove, and replant oak trees. An effective strategy for limiting the spread of invasive forest pathogens like oak wilt is to find and remove diseased trees before the infection spreads to nearby trees. By using hyperspectral imagery to identify oak wilt in forest land, researchers are on their way to helping forest managers identify the best places to treat before the disease can spread throughout the system.